Tumor Site- and Stage-Specific Associations between Allelic Variants of Glutathione S-Transferase and DNA-Repair Genes and Overall Survival in Colorectal Cancer Patients Receiving 5-Fluorouracil-Based Chemotherapy

Introduction Our retrospective cohort study investigated the effect of tumor site and stage on the associations between the allelic variants of glutathione S-transferase (GST) and DNA-repair genes and overall survival (OS) in CRC patients treated with 5-fluorouracil (5-FU)-based adjuvant chemotherapy. Material and Methods We genotyped GSTM1, GSTT1, GSTP1 Ile105Val, XRCC1 Arg399Gln, XRCC3 Thr241Met, and XPD Lys751Gln in 491 CRC patients between 1995 and 2001. A Cox proportional-hazards model was used to calculate the hazard ratios (HRs) and 95% confidence intervals (CIs) for the relationships between the allelic variants and OS. Survival analyses were performed for each allelic variant by using the log-rank test and Kaplan-Meier analysis. Results The CRC patients with the XPD Gln allelic variants had poorer survival than patients with the Lys/Lys genotype (HR  = 1.38, 95% CI  = 1.02–1.87), and rectal cancer patients had the poorest survival among them (HR  = 1.87, 95% CI  = 1.18–2.95). A significantly shorter OS was observed among stage II/III colon cancer patients with the XRCC1 Gln allelic variants (HR  = 1.69, 95% CI  = 1.06–2.71), compared to those with XRCC1 Arg/Arg genotype. In the combined analysis of the XRCC1 and XPD genes patients with stage II/III tumors, the poorest OS occurred in colon cancer patients with the XRCC1 Gln and XPD Gln allelic variants (HR  = 2.60, 95% CI  = 1.19–5.71) and rectal cancer patients with the XRCC1 Arg/Arg and XPD Gln allelic variants (HR  = 2.77, 95% CI  = 1.25–6.17). Conclusion The XPD and XRCC1 allelic variants may be prognostic markers for CRC patients receiving 5-FU based chemotherapy. The contributions of the XPD and XRCC1 allelic variants to OS are tumor site- and/or stage-dependent.


Introduction
After surgery, most metastatic colorectal cancer (CRC) patients receive an adjuvant chemotherapy regimen consisting of combination therapy using 5-fluorouracil (FU) and oxaliplatin (FOL-FOX) or leucovorin (LV) for 6 to 8 mo to reduce the probability of tumor recurrence and prolong survival [1][2][3]. The functional single-nucleotide polymorphisms (SNPs) in drug-targeted genes, including xenobiotic-metabolizing and DNA-repair genes, correlate with variability in clinical outcome in multiple types of cancer [4][5][6][7][8]. The identification of genetic markers may help identify patients who may benefit from chemotherapy and reduce potential toxicity.
Variable chemosensitivity may involve the detoxification pathway, including the glutathione S-transferases (GSTs). The GSTs are multigene families of enzymes that inactivate electrophilic xenobiotics by conjugation with glutathione, preventing DNA damage and adduct formation [9]. Interindividual differences in GST activity may be mediated by genetic polymorphisms [9][10][11][12]. The structural deletion polymorphisms in GSTM1 and GSTT1 result in the loss of enzyme-catalyzed detoxification activity [10], and are predictors of clinical outcome in gastric cancer patients receiving platinum/5-FU chemotherapy [13]. In addition, the reduced glutathione conjugation resulting from polymorphism in the GSTP1 Ile105Val (rs1695) coding region may be associated with an increased survival in CRC patients treated with oxaliplatin/5-FU [14,15].
Various DNA-repair enzymes play important roles in preventing treatment resistance and protecting the genome against carcinogenesis [16][17][18][19]. The expression of the base excision repair (BER) gene family is triggered by internal oxidative stress and DNA damage [20]. The BER pathway involves the X-ray crosscomplementing group 1 gene XRCC1 [21]. The XRCC1 protein directly associates with polymerase beta, DNA ligase III, and poly (ADP-ribose) polymerase (PARP) in single-strand break-repair processes that may play a role in tumor cell sensitivity to 5-FU treatment [22,23]. The XRCC3 protein, a member of the Rad51related enzyme family, contributes to the maintenance of chromosome stability through DNA double-strand break/recombination repair in homologous recombination [24].
The nucleotide excision repair (NER) gene family functions within a range of structurally unrelated DNA lesions and DNA adducts [16,17]. The xeroderma pigmentosum complementation group D (XPD) protein, a member of the NER family functions in sensing, binding, and the subsequent recruitment of repair-related factors [5]. Previous studies have shown that the allelic variants XRCC1 Arg399Gln (rs25487), XRCC3 Thr241Met (rs861539), and XPD Lys751Gln (rs13181) are associated with DNA adduct levels and repair capacity [5,25,26]. Therefore, these polymorphisms of DNA-repair genes may affect clinical outcomes in cancer patients receiving chemotherapy [7,13,20,27,28].
Our previous studies have shown that the null genotypes of GSTM1 and GSTT1, GSTP1 Ile105Val, XRCC1 Arg399Gln, XRCC3 Thr241Met, and XPD Lys751Gln allelic variants are associated with significantly increased risks of CRC [29][30][31][32]. However, other previous reports of associations between GST and DNA-repair allelic variants and clinical outcomes in CRC have been conflicting [6,11,[33][34][35]. Whether these conflicting findings have been due to tissue-specific differences in gene expression between colonic and rectal tumors is unclear [36] because reports of stratified analyses based on the site of the CRC tumor are scant. Therefore, in our present study, we investigated the relationship between the allelic variants of relevant GST and DNA-repair genes and the chemotherapeutic outcomes in a retrospective CRC cohort who received adjuvant chemotherapy in Taiwan to determine whether differences exist among the associations in the tumor site or pathological stage.

Participant selection
Our study was approved by the Institutional Review Board of China Medical University Hospital and Chang Gung Memorial Hospital (CGMH). All patients provided written informed consent before participation in our study. We reviewed 2716 newly diagnosed and histologically confirmed CRC patients who underwent surgery at CGMH between January 1995 and December 2001. We enrolled 499 patients without familial adenomatous polyposis or hereditary nonpolyposis colorectal cancer who underwent 5-FU-based adjuvant chemotherapy as first-line treatment after surgery, and received follow-up examinations every 3 to 6 mo at the outpatient clinic of the Colorectal Section of CGMH.
The postoperative adjuvant chemotherapy regimens included both daily and weekly or monthly treatments. In the daily treatment, tegafur (300-350 mg/m 2 /d) and levamisole (45 mg/d) were administered orally in 3 doses per day for 28 d, followed by a 7-d rest period, totaling 12 mo. The weekly treatments consisted of one of the following treatment schemes for 6 wk, followed by a 2 wk rest, totaling 12 mo: (1) 5-FU (450 mg/m 2 /d) and leucovorin (50 mg/d) were administered as an intravenous bolus once weekly; (2) 5-FU (450 mg/m 2 /d) were administered intravenously once weekly, and levamisole (50 mg) was administered orally 3 times per day for 3 d every 2 wk; or (3) a 24-h intravenous infusion of 5-FU (2600 mg/m 2 /d) and leucovorin (150 mg/d) was administered once weekly. The monthly treatments consisted of one of the following treatment schemes for 6 mo: (1) a continuous 5 d infusion of 5-FU (800 mg/m 2 /d) and leucovorin (50 mg/d) was administered once monthly, or (2) patients received a 5 d continuous infusion of 5-FU (800 mg/ m 2 /d) once monthly combined with levamisole (50 mg) administered orally 3 times per day for 3 d every 2 wk.

Clinical and questionnaire assessment
The clinical data included a medical history, physical examination (including a rectal and perineal examination), and the measurement of carcinoembryonic antigen (CEA). Liver sonograms, chest X-ray, and colonoscopy (or barium enema examination) were performed annually. All patients were followed until May 2008, with a median follow-up time of 48.8 mo (range = 1.5 to 133.3 mo). Overall survival (OS) was measured from the date of surgery until death from any cause. In the calculation of OS, the event date represented the date of death, and the censoring date represented the last known date on which the patient was known to be alive.

DNA extraction and genotyping
The details of the genotyping methods have been described elsewhere [29,30]. DNA was isolated from peripheral blood leukocytes by using sodium dodecyl sulfate (SDS)-proteinase K-RNase digestion and phenol-chloroform extraction. The GSTM1 and GSTT1 genetic variants were identified using polymerase chain reaction (PCR), and the GSTP1 Ile105Val, XRCC1 Arg399Gln, XRCC3 Thr241Met, and XPD Lys751Gln allelic variants were analyzed using PCR combined with restriction fragment length polymorphism.

Statistical analysis
We used x 2 -tests to evaluate possible associations between the categorical variables, and continuous variables were analyzed using a Student t test. Allelic frequencies were compared with those expected at the Hardy-Weinberg equilibrium (HWE) by using a x 2 test. Univariate analysis of the Kaplan-Meier estimates and the logrank test were used to compare the overall survival curves. A Cox proportional-hazards model was used to estimate the hazard ratios (HRs) and the 95% confidence intervals (CIs) for the various allelic variants associated with OS. Age, sex, and the tumor, node, metastasis (TNM) stage were considered potential confounding covariates, and were thus included in the multivariate regression. The Cox proportional-hazards regression model was based on a priori knowledge of factors known to carry prognostic or predictive information to estimate adjusted HRs and 95% CIs. The HRs for differences in OS based on the tumor site and stage were estimated using stratified analysis. All models were examined for adherence to the proportional-hazards assumption by assessing the log-minus-log survival plots and performing the Schoenfeld test. The log-minuslog survival plots and the results of the Schoenfeld tests (P range = 0.22 to 0.90) indicated no violations of the proportionality assumption for the 6 SNPs investigated. All analyses were performed using the SAS statistical software package, version 9.3 for Windows (SAS Institute, Cary, NC, USA). The results of comparisons with a 2 sided P value less than 0.05 were considered to represent statistically significant relationships. To account for multiple comparisons, we calculated the false discovery rate (FDR) by using the PROC MULTTEST in SAS. The statistical power of our analyses was estimated using the PASS statistical software package, version 11 (NCSS, Kaysville, UT, USA).

Demographic and clinical characteristics
Eight CRC patients who failed genotyping were excluded from further analysis. The demographic and clinical characteristics of the 491 patients are summarized in Table 1. The mean age and standard deviation (SD) was 58.5612.5 y. Our cohort consisted of 260 (52.9%) men, and 168 (35.5%) patients had a family history of cancer. The primary tumor was located in the colon in 283 (57.6%) patients and in the rectum in 208 (42.4%) patients. Multiple primary malignancies occurred in 4.9% of the cohort, and 78.4% had a moderately differentiated tumor. The CEA was present at $5 ng/mL in 53.6% of the cohort. In the TNM evaluation, 74 (15.1%), 267 (54.4%), and 150 (30.5%) patients were classified as being stage II, III, and IV, respectively. Compared to patients with rectal cancer, colon cancer patients were younger, were diagnosed more frequently with multiple malignancies, presented more often with moderately differentiated tumors, and were less likely to be TNM stage III (P,0.05). No significant associations were found between sex, family history of cancer, or CEA level and the tumor site.

Genotype frequencies
The genotype distributions of the GST and DNA-repair allelic variants in the CRC patients are also shown in Table 1. The frequencies of the null genotypes of GSTM1 and GSTT1 and the GSTP1 Val, XRCC1 Gln, XRCC3 Met, and XPD Gln allelic

Association between GSTs and DNA-repair allelic variants and survival
Univariate analysis revealed that the association between the XRCC1 Gln allelic variants and reduced OS was more significant than that of XRCC1 Arg/Arg genotype, with a median survival of 43.0 versus 57.1 mo, respectively (HR = 1.36, 95% CI = 1.07-1.73). However, this association was absent following adjustment for the covariates, which included age, sex, and the stage of disease. The other allelic variants analyzed were not associated with OS in univariate analysis of the Kaplan-Meier estimates (data not shown). Table 2 shows the adjusted associations between the allelic variants and OS in the multivariate Cox proportional-hazards regression model. Among the 6 allelic variants tracked, reduced OS was associated with heterozygous carriers of XPD Lys/Gln (HR = 1.34, 95% CI = 0.99-1.84, P = 0.062, FDR = 0.372) and homozygous carriers of XPD Gln/Gln (HR = 2.38, 95% CI = 0.75-7.53, P = 0.141). Compared to patients with XPD Lys/Lys

Association between the genotype and survival by tumor site and TNM stage
The results of the stratified analyses of OS associated with the GSTs and DNA-repair allelic variants based on the tumor location are shown in Table 3. Rectal cancer patients with the XPD Gln allelic variants had a shorter OS compared with patients with XPD Lys/Lys genotype, with a median survival 44.9 versus 51.6 mo (log-rank test P = 0.230). The multivariate Cox regression model also showed that the XPD Gln allelic variants were associated with an 87% increased risk of shorter OS compared to XPD Lys/Lys genotype (HR = 1.87; 95% CI = 1.18-2.95, P = 0.007, FDR = 0.084). However, no adverse effect of Gln allelic variants on OS was observed in colon cancer patients. No significant association between the other allelic variants and OS was observed based on the tumor site.
The stratified analyses of the associations between OS and the polymorphisms of the GST and DNA-repair alleles based on the TNM stage showed that both stage II/III and stage IV CRC patients with the XPD Gln allelic variants didn't have a significantly poorer OS, with HRs of 1.43 (P = 0.085) and 1.31 (P = 0.264), respectively, compared to patients with XPD Lys/Lys genotype ( Table 4). The association between the XRCC1 Gln allelic variants and reduced OS in stage II/III patients approached significance (HR = 1.41, 95% CI = 1.00-1.99, P = 0.051, FDR = 0.510). The other allelic variants were not associated with OS in stage IV patients. Table 5 shows the associations of the XRCC1 and XPD allelic variants with OS in CRC patients stratified by tumor site and TNM stage. The poorest OS was observed among the stage II/III colon cancer patients with the XRCC1 Gln allelic variants (HR = 1.69, 95% CI = 1.06-2.71, P = 0.028, FDR = 0.252). Although the rectal cancer patients who inherited XPD Gln allelic variants had significantly reduced OS, this negative effect was not more prominent in any specific TNM stage (stage II/III, HR = 1.57, 95% CI = 0.88-2.81; stage IV, HR = 1.98, 95% CI = 0.94-4.17).
In the combined analysis of the XRCC1 and XPD allelic variants, the Kaplan-Meier survival curves showed differences in OS among the 4 allelic variants analyzed in stage II/III colon cancer patients (log-rank test P = 0.087, Figure 1a). Compared to patients with XRCC1 Arg/Arg and XPD Lys/Lys genotype, those with XRCC1 Gln and XPD Gln allelic variants had poorer OS (log-rank test P = 0.015), with an HR of 2.60 (95% CI = 1.19-5.71, FDR = 0.102, Table 6). Kaplan-Meier analysis showed significant variability in survival (log-rank test P = 0.021) among the subgroups in stage II/III rectal cancer patients (Figure 1b). The XRCC1 Arg/Arg and XPD Gln allelic variants were associated with a significantly poorer OS, compared with that of patients with XRCC1 Arg/Arg and XPD Lys/Lys genotype (logrank test P = 0.001; HR = 2.77, 95% CI, 1.25-6.17, FDR = 0.102). However, this effect on OS was not observed among stage IV colon or rectal cancer patients with the XRCC1 Arg/Arg and XPD Gln allelic variants. FDR analysis indicated that the

Discussion
We found a significant relationship between the XPD Lys751Gln allelic variant and OS for CRC patients, particularly for rectal cancer patients, whereas XRCC1 Arg399Gln Gln allele correlated with reduced OS in stage II/III colon cancer patients. In addition, the poorest OS was present among the stage II/III colon cancer patients with both XRCC1 Gln and XPD Gln allelic variants and among the stage II/III rectal cancer patients with  Table 5. Tumor site-and tumor-node-metastasis stage-specific hazard ratios for the associations between the XRCC1 Arg399Gln and XPD Lys751Gln allelic variants and overall survival among colorectal cancer patients.  both the XRCC1 Arg/Arg and XPD Gln allelic variants. However, significant associations were not observed between OS and the other GST and DNA-repair gene allelic variants in CRC patients receiving 5-FU-based chemotherapy. The XPD gene is an important DNA-repair gene that codes for enzymes that recruit NER in the repair of a wide range of DNA lesions [16,17]. Mutations that alter the amino acid sequence impact the interactions of XPD enzyme with other members, such as XPA, ERCC1, and replication protein A of the NER complex, resulting in different DNA-repair activities [18]. The XPD Lys751Gln polymorphism converts the basic amino acid, Lys, to the polar amino acid, Gln, at approximately 50 bp upstream from the poly(A) signal, which may affect the function of the XPD protein [4]. Patients with the XPD Lys/Lys genotype have suboptimal DNA-repair activity, and are more sensitive to chemotherapy [5,6].
Previous studies have shown that the XPD Lys751Gln allelic variant may be associated with various clinical outcomes in CRC patients receiving chemotherapy [6,24,34,37]. Park et al [6] found that advanced-stage CRC patients with XPD Gln/Gln genotype treated with chemotherapy tended to have progressive disease and significantly reduced survival, compared to patients with XPD Lys allelic variants (P = 0.002). A previous study of XPD Lys751Gln in 166 advanced CRC patients receiving FOLFOX therapy found a negative relationship between XPD Gln/Gln genotype and adverse progression-free survival (HR = 2.21, 95% CI = 1.17-4.17, P = 0.01) [34]. Our results are consistent with previous studies that have shown a negative association between the XPD Gln allelic variants and OS. However, 2 studies have reported no significant relationship between XPD Lys751Gln and clinical outcomes in CRC patients [7,33].
We demonstrated that the reduction in OS associated with XPD Lys751Gln Gln allelic variants were improved in rectal cancer patients. In the contrast, Duldulao et al. [38] observed that XPD Lys/Lys genotype was significantly associated with increased toxicity to neoadjuvant chemoradiotherapy in 132 stage II/III rectal cancer patients in the United States. However, Cecchin et al [39] reported no association between the XPD Lys751Gln allelic variant and the tumor regression grade in 238 rectal cancer patients treated with neoadjuvant chemoradiotherapy in Italy. Different ethnicities, treatment regiments, and outcome measurements may have contributed to these inconsistencies.
The association between the XRCC1 Arg399Gln allelic variant and OS has been previously reported [35,37]. Artac et al [37] recruited 43 metastatic CRC patients who received irinotecanbased therapy, and found that OS is associated with XRCC1 Gln/ Gln genotype (HR = 2.85, P = 0.04). In a study of stage II/III rectal cancer patients in Spain, patients with XRCC1 Arg/Arg had a greater probability of a positive response to chemoradiotherapy than those with XRCC1 Arg/Gln (OR = 4.18, 95% CI = 1.62-10.7) [35]. In our study, we observed that only the XRCC1 Gln allelic variants were significantly associated with reduced OS in stage II/III colon cancer patients.
In our analysis of the associations between the allelic variants of GST and DNA-repair genes and the OS stratified by tumor site and stage, we found that the XPD 751Gln and the XRCC1 399Gln allelic variants were significantly associated with reduced OS for stage II/III rectal cancer and colon cancer, respectively. Although previous studies have not shown that specific DNA repair activity may vary between the colon and the rectum, a high proportion of rectal tumors has been shown to have reduced levels of thymidylate synthase, based on enzyme activity assays [44,45]. Despite the relatively small number of cases and the lack of statistical significance for the results, our findings support an association between polymorphisms in DNA-repair genes and OS based on specific tumor locations among CRC patients treated with 5-FU chemotherapy. The prognostic value of variability in DNA-repair activity based on a CRC tumor site warrants further study.
Similar to the findings of previous studies, we found no association between OS and the XRCC3 Thr241Met, GSTM1, GSTT1, and GSTP1 Ile105Val allelic variants [14,35,39,46]. A meta-analysis of 13 independent studies that included 1234 advanced or metastatic CRC patients found no significant association between the GSTP1 Ile105Val allelic variant and tumor response [46]. However, Stoehlmacher et al [11] observed Table 6. Tumor site-and tumor-node-metastasis stage-specific hazard ratios for the associations between the coexisting XRCC1 Arg399Gln and XPD Lys751Gln allelic variants and overall survival among colorectal cancer patients. that GSTP1 Val/Val genotype was associated with increased survival in CRC patients following combination therapy with oxaliplatin and 5-FU. The association between the GSTP1 Ile105Val allelic variant and the chemotherapeutic outcome in CRC also warrants further study. There are potential limitations to our findings. First, all of our participants received 5-FU-based chemotherapy. Thus, we did not investigate the genotypes associated with clinical outcomes in an untreated control group. Second, we did not examine mRNA expression or perform immunohistochemistry analysis of tumor tissues. Therefore, our analysis was not free of potential biases, and does not account for the loss of heterozygosity in the tumor. However, immunohistochemistry is a subjectively semiquantitative method, and is limited by the sensitivity of the antibodies and tissue-handling techniques used. Third, our study design lacked statistical power to support the significance of the association between the XPD Ly751Gln and XRCC1 Arg399Gln allelic variants and the OS in CRC patients treated with 5-FU-based chemotherapy (49% and 63%, respectively). Nevertheless, our study included newly diagnosed and histologically confirmed CRC patients, a relatively large sample, and a long-term follow-up. Thus, our results may shed light on the value of the XPD Ly751Gln and XRCC1 Arg399Gln allelic variants as prognostic markers for CRC of various tumor sites and stages.
In conclusion, we evaluated multiple xenobiotic-metabolizing and DNA-repair genetic polymorphisms as prognosticators of OS in CRC patients receiving 5-FU-based chemotherapy. Our results showed that the XPD and XRCC1 allelic variants were tumor site-and/or stage-dependently associated with OS for CRC patients receiving 5-FU-based chemotherapy. Further studies are warranted to identify the underlying biochemical mechanisms affected by the mutations in the allelic variants, and to validate the roles of XPD and XRCC1 genetic polymorphisms as predictors of chemotherapeutic outcome in CRC patients.